Celestica: Hyperscaler Capex May Cool, But Revenue Momentum Still Looks Explosive

Celestica's hyperscaler-driven CCS growth is accelerating, supported by 1.6T networking wins and AI/ML compute ramps, with management guiding strong enterprise growth into 2027. The company is committing about $1 billion of capex this year and at least $1.5 billion next year to expand capacity, which supports long-term demand visibility but may दब pressure near-term free cash flow margins. Downward revisions to hyperscaler capex over the next four quarters could cap near-term upside if customer spending does not reaccelerate.

Analysis

The market is likely underestimating the second-order winner set here: not just CLS, but the broader AI infrastructure stack that benefits from hyperscaler diversification into networking and integration-heavy buildouts. The key signal is that spend is shifting from pure GPU procurement toward more systems-level deployment, which tends to favor vendors with manufacturing depth, supply-chain control, and ability to absorb custom ramps. That usually compresses the moat for smaller networking specialists while widening it for scaled outsourced hardware partners that can take share in platform-level programs.

The near-term tension is cash flow versus backlog visibility. Heavy capex now is a deliberate call option on 2026-2027 demand, but if customer capex trajectories stay merely flat rather than re-accelerating, the market may punish the stock for negative FCF optics before the revenue base fully catches up. That creates a classic “good growth, bad duration” setup: earnings estimates can keep moving higher while multiple expansion stalls because investors remain skeptical that incremental spend will translate into durable margin expansion.

The biggest contrarian risk is that the consensus is treating AI infrastructure demand as linear when it is probably lumpy and allocation-driven. A small number of hyperscalers can defer or re-phase orders for a quarter or two without signaling demand destruction, which is enough to create visible disappointment in high-beta hardware names. If hyperscaler capex revisions keep drifting lower over the next 1-2 quarters, CLS can still win share operationally while the stock de-rates on slower-than-expected conversion of capacity into cash flow.

Key Decisions for Investors

• Long CLS on pullbacks over the next 2-6 weeks, but size it as a “growth with FCF lag” position; target a 12-18 month hold where margin leverage can reassert, and use a 15-20% trailing stop if hyperscaler capex data rolls over.
• Pair trade: long CLS / short a higher-multiple AI networking beneficiary with less manufacturing leverage over the next 3-9 months; the thesis is that outsourced execution and custom integration should capture more of the next deployment wave than pure-play component exposure.
• Buy medium-dated CLS call spreads into any post-earnings or capex-announcement weakness; structure for 6-9 months out to express upside to 2026 guidance while limiting drawdown from near-term free-cash-flow pressure.
• If hyperscaler capex revisions continue down for another quarter, reduce CLS exposure and rotate into beneficiaries with clearer 2025 cash conversion; the risk/reward on CLS weakens if revenue visibility remains intact but the market stops paying for it.
• Watch for evidence of enterprise contribution inflecting in 2H; if that mix shift shows up, CLS could earn a higher multiple even before FCF normalizes, making it one of the better 2025-2027 compounding stories in the hardware space.